Abstract:
Predicting students success has been a very popular study across different fields and with this study we will be focusing on how Machine Learning can aid us in giving us insight in how students will perform in their exams. In this paper, we propose a model that takes in Demographic, Academic and Behavioural attributes and investigate the how these attributes contribute a students performance and also predict at risk students. For the analysis of the features we will use the Mutual Information algorithm, alongside five machine learning models that are a mixture of classification and regression classifiers. We will also use these models to predict our students performance. The five classifiers used in this study are as follows: Gaussian Naive Bayes, Support Vector Machine, Random Forest, K-Nearest Neighbour and Logistic Regression and we achieved prediction accuracy of 50.83%, 81.67%, 78.33%, 75.00% and 74.17% respectively. The results yielded revealed that there is a strong correlation between a students behavioural characteristics and their academic performance.
